Enum FederatedAggregationStrategy

Namespace

AiDotNet.Models.Options

Assembly

AiDotNet.dll

Specifies which federated aggregation strategy to use.

public enum FederatedAggregationStrategy

Fields

Bulyan = 9

Bulyan (Multi-Krum selection + trimming).

FedAvg = 0

Federated Averaging (FedAvg).

FedBN = 2

Federated Batch Normalization (FedBN).

FedProx = 1

Federated Proximal (FedProx) for heterogeneity.

Krum = 7

Krum (Byzantine-robust selection).

Median = 3

Coordinate-wise median aggregation.

MultiKrum = 8

Multi-Krum (select m central updates, then average).

Rfa = 6

Robust Federated Aggregation (geometric median / RFA).

TrimmedMean = 4

Coordinate-wise trimmed mean aggregation.

WinsorizedMean = 5

Coordinate-wise winsorized mean aggregation.

Remarks

For Beginners: In federated learning, each client trains locally and sends an update. The server then combines those updates using an aggregation strategy.